Identification Techniques for Large Structures

This research activity will utilize the least-squares identification technique for tuning the dynamic models used for the controller design to achieve precision pointing. System identification from measured frequency responses involves fitting the parameterized transfer function matrix of a theoretical linear time-invariant system to a measured transfer function matrix at selected frequencies. A criterion is needed to qualify the error between the theoretical and the measured responses at the test frequencies. The least-squares identification technique does not require any a priori error information. It correlates the measured data at the test frequencies to minimize an average of modeling errors. Use of the least-squares method makes it easy to compute the cost function of the identification error, and to optimize with relative ease. These issues are critical when dealing with large complex systems.

An actively controlled structure such as a segmented reflector telescope involves a large number of sensors and actuators, increasing significantly the likelihood of a sensor/actuator failure. To address the substantial performance degradation resulting from such a failure, a task will be included to study suitable methodologies for failure detection, fault location and isolation, and reconfiguration of the control system for performance recovery.

All control algorithms developed in this research will be experimentally validated on the SPACE testbed. As with the other research tasks, the progress and results of the technical activities will be formally documented in annual reports and a final comprehensive report will be issued at the completion of the program in the fifth year.